Cyclostationnarity analysis of Enso memory. Part of the INC ("Incendies Nouvelle Caledonie") ANR project: Improvement of fire management for biodiversity conservation in New Caledonia by developing a comprehensive understanding of the climatic, ecological and social drivers of fire
Résumé
Most climate fluctuations may be modulated by a variety of periodic or quasi-periodic deterministic forcing (e.g. diurnal, seasonal, Milankovitch cycle). This process modulation often induce cyclostationary (CS) behaviour defined as periodic correlations. We proposed here that quasi-periodic modulations may appear when the inertia (memory) of a process is modulating its pure CS property. Such process was said to exhibit an Extended AutoCorrelation (EAC) if its time-variant autocorrelation was periodic 'in average' in time and frequency domains. A new statistical tool has been developed to analyze EAC in a time series with the help of a comfortable time-frequency visualization, called Digital X Model (DXM). We examined the seasonal-to-interannual varibility of El Niño/Southern Oscillation (ENSO) signals by the DXM. Results were compared to those obtained by the Monte-Carlo Singular Spectrum Analysis (SSA), the wavelet analysis and the Empirical Mode Decomposition (EMD). This comparative approach applied to SOI and SST (Niño3) indices confirmed the well-know interannual periodicities discussed in literature: the quasi-biennial (QB), 29 months, and the quasi-quadrennial (QQ), 50 months. The main result presented here was that these significant cycles were the only one detected by the four methods, suggesting that the QB and the QQ would be the fundamental modes of ENSO. In addition, the QB was strongly associated to the atmospheric component of ENSO and the QQ associated to the oceanic component of ENSO. This results lead to the hypothesis that the QB and QQ could be separate oscillations generated by differing physics. The interaction of theses fundamental modes with the seasonal cycle could produce three additional minor periodicities by combination of tones, the 15-, 20-and 69 months oscillations, detected by the DXM and sometimes found in the three other methods.